def decode_test(session,
decode_op,
network,
dataset,
label_type,
save_path=None):
"""Visualize label outputs.
Args:
session: session of training model
decode_op: operation for decoding
network: network to evaluate
dataset: An instance of a `Dataset` class
label_type: stirng, phone39 or phone48 or phone61 or character
save_path: path to save decoding results
"""
# Batch size is expected to be 1
iteration = dataset.data_num
# Make data generator
mini_batch = dataset.next_batch(batch_size=1)
if label_type == 'character':
map_file_path = '../metrics/mapping_files/attention/char2num.txt'
else:
map_file_path = '../metrics/mapping_files/attention/phone2num_' + \
label_type[5:7] + '.txt'
# if save_path is not None:
# sys.stdout = open(join(network.model_dir, 'decode.txt'), 'w')
for step in range(iteration):
# Create feed dictionary for next mini batch
inputs, labels_true, inputs_seq_len, _, input_names = mini_batch.__next__(
)
feed_dict = {
network.inputs: inputs,
network.labels: labels_true,
network.inputs_seq_len: inputs_seq_len,
network.keep_prob_input: 1.0,
network.keep_prob_hidden: 1.0
}
# Visualize
labels_pred = session.run(decode_op, feed_dict=feed_dict)
if label_type == 'character':
print('----- wav: %s -----' % input_names[0])
print('True: %s' % num2char(labels_true[0][1:-1], map_file_path))
print('Pred: %s' %
num2char(labels_pred[0], map_file_path).replace('>', ''))
else:
print('----- wav: %s -----' % input_names[0])
print('True: %s' % num2phone(labels_true[0][1:-1], map_file_path))
print('Pred: %s' %
num2phone(labels_pred[0], map_file_path).replace('>', ''))
def decode_test(session, decode_op, network, dataset, label_type,
save_path=None):
"""Visualize label outputs of CTC model.
Args:
session: session of training model
decode_op: operation for decoding
network: network to evaluate
dataset: An instance of a `Dataset` class
label_type: string, kanji or kana or phone
save_path: path to save decoding results
"""
# Batch size is expected to be 1
iteration = dataset.data_num
# Make data generator
mini_batch = dataset.next_batch(batch_size=1)
if label_type == 'kanji':
map_file_path = '../metrics/mapping_files/ctc/kanji2num.txt'
elif label_type == 'kana':
map_file_path = '../metrics/mapping_files/ctc/kana2num.txt'
elif label_type == 'phone':
map_file_path = '../metrics/mapping_files/ctc/phone2num.txt'
if save_path is not None:
sys.stdout = open(join(network.model_dir, 'decode.txt'), 'w')
for step in range(iteration):
# Create feed dictionary for next mini batch
inputs, labels_true, inputs_seq_len, input_names = mini_batch.__next__()
# NOTE: labels_true is expected to be a list of string when evaluation
# using dataset where label_type is kanji or kana
feed_dict = {
network.inputs: inputs,
network.inputs_seq_len: inputs_seq_len,
network.keep_prob_input: 1.0,
network.keep_prob_hidden: 1.0
}
# Visualize
labels_pred_st = session.run(decode_op, feed_dict=feed_dict)
labels_pred = sparsetensor2list(labels_pred_st, batch_size=1)
if label_type in ['kanji', 'kana']:
print('----- wav: %s -----' % input_names[0])
print('True: %s' % labels_true[0])
print('Pred: %s' % num2char(labels_pred[0], map_file_path))
elif label_type == 'phone':
print('----- wav: %s -----' % input_names[0])
print('True: %s' % num2phone(labels_true[0], map_file_path))
print('Pred: %s' % num2phone(labels_pred[0], map_file_path))
def check_loading(self, num_gpu, is_sorted):
print('----- num_gpu: ' + str(num_gpu) + ', is_sorted: ' +
str(is_sorted) + ' -----')
batch_size = 64
dataset = Dataset(data_type='train',
label_type_main='character',
label_type_sub='phone61',
batch_size=batch_size,
num_stack=3,
num_skip=3,
is_sorted=is_sorted,
is_progressbar=True,
num_gpu=num_gpu)
tf.reset_default_graph()
with tf.Session().as_default() as sess:
print('=> Loading mini-batch...')
map_file_path_char = '../metrics/mapping_files/ctc/char2num.txt'
map_file_path_phone = '../metrics/mapping_files/ctc/phone2num_61.txt'
mini_batch = dataset.next_batch(session=sess)
iter_per_epoch = int(dataset.data_num / (batch_size * num_gpu)) + 1
for i in range(iter_per_epoch + 1):
return_tuple = mini_batch.__next__()
inputs = return_tuple[0]
labels_char_st = return_tuple[1]
labels_phone_st = return_tuple[2]
if num_gpu > 1:
for inputs_gpu in inputs:
print(inputs_gpu.shape)
labels_char_st = labels_char_st[0]
labels_phone_st = labels_phone_st[0]
labels_char = sparsetensor2list(labels_char_st,
batch_size=len(inputs))
labels_phone = sparsetensor2list(labels_phone_st,
batch_size=len(inputs))
if num_gpu == 1:
for inputs_i, labels_i in zip(inputs, labels_char):
if len(inputs_i) < len(labels_i):
print(len(inputs_i))
print(len(labels_i))
raise ValueError
for inputs_i, labels_i in zip(inputs, labels_phone):
if len(inputs_i) < len(labels_i):
print(len(inputs_i))
print(len(labels_i))
raise ValueError
str_true_char = num2char(labels_char[0], map_file_path_char)
str_true_char = re.sub(r'_', ' ', str_true_char)
str_true_phone = num2phone(labels_phone[0],
map_file_path_phone)
def decode_test_multitask(session, decode_op_main, decode_op_second, network,
dataset, label_type_second, save_path=None):
"""Visualize label outputs of Multi-task CTC model.
Args:
session: session of training model
decode_op_main: operation for decoding in the main task
decode_op_second: operation for decoding in the second task
network: network to evaluate
dataset: An instance of a `Dataset` class
label_type_second: string, phone39 or phone48 or phone61
save_path: path to save decoding results
"""
# Batch size is expected to be 1
iteration = dataset.data_num
# Make data generator
mini_batch = dataset.next_batch(batch_size=1)
if save_path is not None:
sys.stdout = open(join(network.model_dir, 'decode.txt'), 'w')
# Decode character
print('===== character =====')
map_file_path = '../metrics/mapping_files/ctc/char2num.txt'
for step in range(iteration):
# Create feed dictionary for next mini batch
inputs, labels_true_st, _, inputs_seq_len, input_names = mini_batch.__next__()
feed_dict = {
network.inputs: inputs,
network.inputs_seq_len: inputs_seq_len,
network.keep_prob_input: 1.0,
network.keep_prob_hidden: 1.0
}
# Visualize
labels_pred_st = session.run(decode_op_main, feed_dict=feed_dict)
labels_true = sparsetensor2list(labels_true_st, batch_size=1)
labels_pred = sparsetensor2list(labels_pred_st, batch_size=1)
print('----- wav: %s -----' % input_names[0])
print('True: %s' % num2char(
labels_true[0], map_file_path))
print('Pred: %s' % num2char(
labels_pred[0], map_file_path))
# Decode phone
print('\n===== phone =====')
map_file_path = '../metrics/mapping_files/ctc/phone2num_' + \
label_type_second[5:7] + '.txt'
for step in range(iteration):
# Create feed dictionary for next mini batch
inputs, _, labels_true_st, inputs_seq_len, input_names = mini_batch.__next__()
feed_dict = {
network.inputs: inputs,
network.inputs_seq_len: inputs_seq_len,
network.keep_prob_input: 1.0,
network.keep_prob_hidden: 1.0
}
# Visualize
labels_pred_st = session.run(decode_op_second, feed_dict=feed_dict)
labels_true = sparsetensor2list(labels_true_st, batch_size=1)
labels_pred = sparsetensor2list(labels_pred_st, batch_size=1)
print('----- wav: %s -----' % input_names[0])
print('True: %s' % num2phone(
labels_true[0], map_file_path))
print('Pred: %s' % num2phone(
labels_pred[0], map_file_path))
def do_eval_per(session,
decode_op,
per_op,
network,
dataset,
label_type,
eos_index,
eval_batch_size=None,
is_progressbar=False):
"""Evaluate trained model by Phone Error Rate.
Args:
session: session of training model
decode_op: operation for decoding
per_op: operation for computing phone error rate
network: network to evaluate
dataset: An instance of a `Dataset' class
label_type: string, phone39 or phone48 or phone61
eos_index: int, the index of <EOS> class
eval_batch_size: int, the batch size when evaluating the model
is_progressbar: if True, visualize the progressbar
Returns:
per_global: An average of PER
"""
if eval_batch_size is not None:
batch_size = eval_batch_size
else:
batch_size = dataset.batch_size
train_label_type = label_type
data_label_type = dataset.label_type
num_examples = dataset.data_num
iteration = int(num_examples / batch_size)
if (num_examples / batch_size) != int(num_examples / batch_size):
iteration += 1
per_global = 0
# Make data generator
mini_batch = dataset.next_batch(batch_size=batch_size)
phone2num_map_file_path = '../metrics/mapping_files/attention/phone2num_' + \
train_label_type[5:7] + '.txt'
phone2num_39_map_file_path = '../metrics/mapping_files/attention/phone2num_39.txt'
phone2phone_map_file_path = '../metrics/mapping_files/phone2phone.txt'
for step in wrap_iterator(range(iteration), is_progressbar):
# Create feed dictionary for next mini-batch
inputs, att_labels_true, _, inputs_seq_len, _, _ = mini_batch.__next__(
)
feed_dict = {
network.inputs: inputs,
network.inputs_seq_len: inputs_seq_len,
network.keep_prob_input: 1.0,
network.keep_prob_hidden: 1.0
}
batch_size_each = len(inputs_seq_len)
if False:
# Evaluate by 61 phones
per_local = session.run(per_op, feed_dict=feed_dict)
per_global += per_local * batch_size_each
else:
# Evaluate by 39 phones
predicted_ids = session.run(decode_op, feed_dict=feed_dict)
predicted_ids_phone39 = []
labels_true_phone39 = []
for i_batch in range(batch_size_each):
# Convert from num to phone (-> list of phone strings)
phone_pred_seq = num2phone(predicted_ids[i_batch],
phone2num_map_file_path)
phone_pred_list = phone_pred_seq.split(' ')
# Mapping to 39 phones (-> list of phone strings)
phone_pred_list = map_to_39phone(phone_pred_list,
train_label_type,
phone2phone_map_file_path)
# Convert from phone to num (-> list of phone indices)
phone_pred_list = phone2num(phone_pred_list,
phone2num_39_map_file_path)
predicted_ids_phone39.append(phone_pred_list)
if data_label_type != 'phone39':
# Convert from num to phone (-> list of phone strings)
phone_true_seq = num2phone(att_labels_true[i_batch],
phone2num_map_file_path)
phone_true_list = phone_true_seq.split(' ')
# Mapping to 39 phones (-> list of phone strings)
phone_true_list = map_to_39phone(
phone_true_list, train_label_type,
phone2phone_map_file_path)
# Convert from phone to num (-> list of phone indices)
phone_true_list = phone2num(phone_true_list,
phone2num_39_map_file_path)
labels_true_phone39.append(phone_true_list)
else:
labels_true_phone39 = att_labels_true
# Compute edit distance
labels_true_st = list2sparsetensor(labels_true_phone39,
padded_value=eos_index)
labels_pred_st = list2sparsetensor(predicted_ids_phone39,
padded_value=eos_index)
per_local = compute_edit_distance(session, labels_true_st,
labels_pred_st)
per_global += per_local * batch_size_each
per_global /= dataset.data_num
return per_global
def do_eval_per(session,
decode_op,
per_op,
network,
dataset,
label_type,
eos_index,
eval_batch_size=None,
is_progressbar=False,
is_multitask=False):
"""Evaluate trained model by Phone Error Rate.
Args:
session: session of training model
decode_op: operation for decoding
per_op: operation for computing phone error rate
network: network to evaluate
dataset: An instance of a `Dataset' class
label_type: string, phone39 or phone48 or phone61
eos_index: int, the index of <EOS> class
eval_batch_size: int, the batch size when evaluating the model
is_progressbar: if True, visualize the progressbar
is_multitask: if True, evaluate the multitask model
Returns:
per_mean: An average of PER
"""
if eval_batch_size is not None:
batch_size = eval_batch_size
else:
batch_size = dataset.batch_size
train_label_type = label_type
eval_label_type = dataset.label_type
num_examples = dataset.data_num
iteration = int(num_examples / batch_size)
if (num_examples / batch_size) != int(num_examples / batch_size):
iteration += 1
per_mean = 0
# Make data generator
mini_batch = dataset.next_batch(batch_size=batch_size)
train_phone2num_map_file_path = '../metrics/mapping_files/ctc/' + \
train_label_type + '_to_num.txt'
eval_phone2num_map_file_path = '../metrics/mapping_files/ctc/' + \
train_label_type + '_to_num.txt'
phone2num_39_map_file_path = '../metrics/mapping_files/ctc/phone39_to_num.txt'
phone2phone_map_file_path = '../metrics/mapping_files/phone2phone.txt'
for step in wrap_iterator(range(iteration), is_progressbar):
# Create feed dictionary for next mini-batch
if not is_multitask:
inputs, labels_true, inputs_seq_len, _, _ = mini_batch.__next__()
else:
inputs, _, labels_true, inputs_seq_len, _, _ = mini_batch.__next__(
)
feed_dict = {
network.inputs: inputs,
network.inputs_seq_len: inputs_seq_len,
network.keep_prob_input: 1.0,
network.keep_prob_hidden: 1.0
}
batch_size_each = len(inputs_seq_len)
# Evaluate by 39 phones
predicted_ids = session.run(decode_op, feed_dict=feed_dict)
labels_pred_mapped, labels_true_mapped = [], []
for i_batch in range(batch_size_each):
###############
# Hypothesis
###############
# Convert from num to phone (-> list of phone strings)
phone_pred_list = num2phone(
predicted_ids[i_batch],
train_phone2num_map_file_path).split(' ')
# Mapping to 39 phones (-> list of phone strings)
phone_pred_list = map_to_39phone(phone_pred_list, train_label_type,
phone2phone_map_file_path)
# Convert from phone to num (-> list of phone indices)
phone_pred_list = phone2num(phone_pred_list,
phone2num_39_map_file_path)
labels_pred_mapped.append(phone_pred_list)
###############
# Reference
###############
# Convert from num to phone (-> list of phone strings)
phone_true_list = num2phone(
labels_true[i_batch], eval_phone2num_map_file_path).split(' ')
# Mapping to 39 phones (-> list of phone strings)
phone_true_list = map_to_39phone(phone_true_list, eval_label_type,
phone2phone_map_file_path)
# Convert from phone to num (-> list of phone indices)
phone_true_list = phone2num(phone_true_list,
phone2num_39_map_file_path)
labels_true_mapped.append(phone_true_list)
# Compute edit distance
labels_true_st = list2sparsetensor(labels_true_mapped,
padded_value=eos_index)
labels_pred_st = list2sparsetensor(labels_pred_mapped,
padded_value=eos_index)
per_each = compute_edit_distance(session, labels_true_st,
labels_pred_st)
per_mean += per_each * batch_size_each
per_mean /= dataset.data_num
return per_mean
